PICBench introduces the first evaluation framework to test how Large Language Models can automate complex photonic chip design tasks.

• Addresses a critical gap in automating the time-consuming and error-prone design process for Photonic Integrated Circuits (PICs)
• Evaluates LLMs on realistic PIC design tasks including code generation, layout optimization, and error correction
• Provides standardized benchmarks to measure LLM effectiveness in this specialized engineering domain
• Enables meaningful comparison across different LLM approaches for photonic chip automation

This research opens new possibilities for accelerating photonic chip development, potentially reducing design time while improving reliability - critical as PICs become increasingly important for next-generation computing and communications.

PICBench: Benchmarking LLMs for Photonic Integrated Circuits Design